Metabolic Syndrome Prediction Models Using Machine Learning and Sasang Constitution Type

Area of Science:

• Computational biology
• Personalized medicine
• Health informatics

Background:

• Machine learning (ML) shows promise for predicting metabolic syndrome (MetS).
• Previous research indicates Sasang constitution type influences MetS risk.
• This study explores ML for MetS prediction and the impact of Sasang type.

Purpose of the Study:

• To develop ML models for predicting MetS.
• To evaluate if incorporating Sasang constitution type enhances MetS prediction accuracy.
• To compare ML models against conventional logistic regression.

Main Methods:

• Recruited 2,871 participants for health check-ups (2005-2006).
• Applied six ML algorithms (k-NN, Naive Bayes, Random Forest, Decision Tree, MLP, SVM) and logistic regression.
• Compared ML models with and without Sasang constitution type, considering factors like BMI, stress, and lifestyle.

Main Results:

• 750 participants (26.1%) had MetS.
• ML models (MLP, SVM) performed comparably to logistic regression (AUC).
• Naive Bayes achieved the highest sensitivity (0.49) vs. logistic regression (0.39).
• Sasang type integration improved sensitivity for most ML models, excluding k-NN.

Conclusions:

• ML models are effective tools for predicting MetS.
• Integrating Sasang constitution type can enhance the sensitivity of ML-based MetS prediction models.